Key points
- Carbon is open to paying developers and enterprises through Google's AI Studio and business offering, and to AI Ultra subscribers in the Gemini app, Search AI Mode and Antigravity.
- Introductory pricing matches Argon, but Carbon spends more tokens per task, so real bills may run higher, according to The Decoder.
- Google is previewing /swarm in Antigravity, which turns one prompt into a team of agents with Carbon handling design, coordination and auditing.
- Google says Carbon ranks first on CursorBench 4.0, WANDR and AutomationBench at over 2X lower cost per task than the next-best external model.
Why it matters for marketers: Marketing teams building agent workflows get a stronger model at launch pricing, but token-heavy reasoning means they should test cost per finished task, not per-token rates.
Google has moved Gemini 4 Carbon from a two-week limited release for cybersecurity defenders to general availability for paying developers and businesses. Ultra subscribers get it in the Gemini app, AI Mode in Search and Antigravity. The Verge notes it is unclear whether lower-tier or free users will get access.
Google positions Carbon as a companion to Gemini 4 Argon, not a replacement. Argon stays the recommended pick for most software and knowledge work, while Carbon targets hard multi-step coding, debugging and open-ended research. Alongside the launch, Google is previewing /swarm in Antigravity, where one prompt becomes a coordinated group of agents you can watch in real time.
For marketers, the cost picture needs care. The Verge says the introductory rate is well below rivals' per-token prices and will double at the end of November. The Decoder points out that Carbon uses more tokens than Argon, so total spend per task is likely higher even at an identical rate. The Verge also cautions that companies can cherry-pick benchmark results.
Competition is tight. Both outlets report that Anthropic's Fable 5.2 launched just before Carbon and edges it on some independent tests, though Google claims Carbon wins on several workflow benchmarks at lower cost. The Decoder adds that the value of agent swarms is still debated, with parallel speed as the main benefit.
Teams should pilot Carbon on one demanding workflow, such as multi-step automation or large document analysis, and compare total cost and output quality against Argon before the introductory pricing ends.
Sources
This brief is our own summary. Read the full reporting at the source.